kimu-counter-consensus-reasoning

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user-provided content such as drafts, outlines, and opinion pieces.\n
  • Ingestion points: User-supplied text provided through the template placeholder 【原文】.\n
  • Boundary markers: The instructions lack explicit delimiters or "ignore instructions" warnings for the ingested data.\n
  • Capability inventory: The skill is limited to text analysis and rewriting within the agent's context and does not use tools for network, filesystem, or shell access.\n
  • Sanitization: There are no defined routines for sanitizing or filtering the input data.\n- [SAFE]: The skill provides detailed analytical guidelines based on communication theories (Luhmann, Carey, Noelle-Neumann) and strictly adheres to logical analysis and content rewriting without requesting elevated privileges or executing remote code.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 02:01 AM
Security Audit — agent-trust-hub — kimu-counter-consensus-reasoning